The image feature extraction method of the present invention includes: the step of performing k2 dividing process at least once on a given image so as to convert the given image into a multi-divided image, where the k2 dividing process comprises the steps of: a) creating matrix T based on image matrix X; b) computing singular values of the matrix T; c) determining whether or not minj|σj−σj−1|>ε; d) if the result of the determination in the step c) is “No”, returning to the step c) subsequent to computing the singular values of the enlarged matrix Tα; e) if the result of the determination in the step c) is “Yes”, obtaining U which satisfies T=USVT; f) obtaining matrix T1=UTT; and g) creating image matrix X1 based on matrix T1.
Legal claims defining the scope of protection, as filed with the USPTO.
1. An image feature extraction method for extracting a feature of an image, comprising the step of performing k 2 (k is an arbitrary integer greater than or equal to 2) dividing process at least once on a given image so as to convert the given image into a multi-divided image, wherein the k 2 dividing process comprises the steps of: a) creating matrix T based on image matrix X; b) computing singular values σ 1 , σ 2 , . . . , σ k^2 of the matrix T, where σ 1 ≧σ 2 ≧ . . . ≧σ k^2 ; c) determining whether or not min j |σ j −σ j−1 |>ε, where ε shows a constant greater than or equal to machine epsilon; d) if the result of the determination in the step c) is “No” returning to the step c) subsequent to computing the singular values of enlarged matrix T α ; e) if the result of the determination in the step c) is “Yes”, obtaining U which satisfies T=USV T , where S=diag (σ 1 , σ 2 , . . . , σ k^2 ), U is an orthogonal matrix and V is an orthogonal matrix; f) obtaining matrix T 1 =U T T; and g) creating image matrix X 1 based on matrix T 1 , wherein the computing the singular values of the enlarged matrix T α in the step d) comprises the steps of: creating enlarged matrix T α based on matrix T and a frame added to at least a portion of at least one side of image matrix X, the frame having a size of at least k pixels and k pixels; and computing singular values σ 1 , σ 2 , . . . , σ k^2 of the enlarged matrix T α , where σ 1 ≧σ 2 ≧ . . . ≧σ k^2 .
2. An image feature extraction method for extracting a feature of an image, comprising the step of performing k 2 (k is an arbitrary integer greater than or equal to 2) dividing process at least once on a given image so as to convert the given image into a multi-divided image, wherein the k 2 dividing process comprises the steps of: a) creating matrix T based on image matrix X; b) obtaining a singular value decomposition of the matrix T, T=USV T , where S=diag (σ 1 , σ 2 , . . . , σ k^2 ), σ 1 , σ 2 , . . . , σ k^2 are singular values of T satisfying σ 1 ≧σ 2 ≧ . . . ≧σ k^2 , U is an orthogonal matrix and V is an orthogonal matrix; c) determining whether or not min j |σ j −σ j−1 |>ε, where ε shows a constant greater than or equal to machine epsilon; d) if the result of the determination in the step c) is “No”, returning to the step c) subsequent to performing singular value decomposition of enlarged matrix T α ; e) if the result of the determination in the step c) is “Yes”, obtaining matrix T 1 =U T T; and f) obtaining matrix T 1 =U T T based on matrix T 1 , wherein the performing singular value decomposition of enlarged matrix T α comprises the steps of: creating enlarged matrix T α based on matrix T and a frame added to at least a portion of at least one side of image matrix X, the frame having a size of at least k pixels and k pixels; and obtaining the singular value decomposition of the enlarged matrix T α , T α =USV T , where S=diag (σ 1 , σ 2 , . . . , σ k^2 ), σ 1 , σ 2 , . . . , σ k^2 is a singular value of T α which satisfies σ 1 ≧σ 2 ≧ . . . ≧σ k^2 , U is an orthogonal matrix, and V is an orthogonal matrix.
3. An image feature extraction method according to claim 1 , wherein the image is a gray scale image or a color image.
4. An image feature extraction method according to claim 2 , wherein the image is a gray scale image or a color image.
5. An image feature extraction method according to claim 1 , wherein the singular value decompositions of the T and the T α are performed by floating point arithmetic.
6. An image feature extraction method according to claim 2 , wherein the singular value decompositions of the T and the T α are performed by floating point arithmetic.
7. An image feature extraction method according to claim 1 , wherein the singular value decompositions of the T and the T α are performed by integer arithmetic.
8. An image feature extraction method according to claim 2 , wherein the singular value decompositions of the T and the T α are performed by integer arithmetic.
9. An image feature extraction method according to claim 1 , wherein a known k 2 dividing process is used together with the k 2 dividing process, so that the given image matrix X is converted into a multi-divided image.
10. An image feature extraction method according to claim 2 , wherein a known k 2 dividing process is used together with the k 2 dividing process, so that the given image matrix X is converted into a multi-divided image.
11. An image compression method for compressing an image, comprising the steps of: performing k 2 (k is an arbitrary integer greater than or equal to 2) dividing process at least once on a given image so as to convert the given image into a multi-divided image; and performing a data compression process on the multi-divided image so as to create a compressed image, wherein the k 2 dividing process comprises the steps of: a) creating matrix T based on image matrix X; b) computing singular values σ 1 , σ 2 , . . . , σ k^2 of the matrix T, where σ 1 ≧σ 2 ≧ . . . ≧σ k^2 ; c) determining whether or not min j |σ j −σ j−1 |>ε, where ε shows a constant greater than or equal to machine epsilon; d) if the result of the determination in the step c) is “No”, returning to the step c) subsequent to computing the singular values of the enlarged matrix T α ; e) if the result of the determination in the step c) is “Yes”, obtaining U which satisfies T=USV T , where S=diag (σ 1 , σ 2 , . . . , σ k^2 ), U is an orthogonal matrix and V is an orthogonal matrix; f) obtaining matrix T 1 =U T T; and g) creating image matrix X 1 based on matrix T 1 , wherein the computing the singular values of the enlarged matrix T α in the step d) comprises the steps of: creating enlarged matrix T α based on matrix T and a frame added to at least a portion of at least one side of image matrix X, the frame having a size of at least k pixels and k pixels; and computing singular values σ 1 , σ 2 , . . . , σ k^2 of the enlarged matrix T α , where σ 1 ≧σ 2 ≧ . . . ≧σ k^2 .
12. An image compression method for compressing an image, comprising the steps of: performing k 2 (k is an arbitrary integer greater than or equal to 2) dividing process at least once on a given image so as to convert the given image into a multi-divided image; and performing a data compression process on the multi-divided image so as to create a compressed image, wherein the k 2 dividing process comprises the steps of: a) creating matrix T based on image matrix X; b) obtaining a singular value decomposition of the matrix T, T=USV T , where S=diag (σ 1 , σ 2 , . . . , σ k^2 ), σ 1 , σ 2 , . . . , σ k^2 are singular values of T satisfying σ 1 ≧σ 2 ≧ . . . ≧σ k^2 , U is an orthogonal matrix and V is an orthogonal matrix; c) determining whether or not min j |σ j −σ j−1 |>ε, where ε shows a constant greater than or equal to machine epsilon; d) if the result of the determination in the step c) is “No”, returning to the step c) subsequent to performing singular value decomposition of enlarged matrix T α ; e) if the result of the determination in the step c) is “Yes”, obtaining matrix T 1 =U T T; and f) obtaining matrix T 1 =U T T based on matrix T 1 , wherein the performing singular value decomposition of enlarged matrix T α comprises the steps of: creating enlarged matrix T α based on matrix T and a frame added to at least a portion of at least one side of image matrix X, the frame having a size of at least k pixels and k pixels; and obtaining the singular value decomposition of the enlarged matrix T α , T α =USV T , where S=diag (σ 1 , σ 2 , . . . , σ k^2 ), σ 1 , σ 2 , . . . , σ k^2 is a singular value of T α which satisfies σ 1 ≧σ 2 ≧ . . . ≧σ k^2 , U is an orthogonal matrix, and V is an orthogonal matrix.
13. An image compression method according to claim 11 , wherein the image is a gray scale image or a color image.
14. An image compression method according to claim 12 , wherein the image is a gray scale image or a color image.
15. An image compression method according to claim 11 , wherein the singular value decompositions of the T and the T α are performed by floating point arithmetic.
16. An image compression method according to claim 12 , wherein the singular value decompositions of the T and the T α are performed by floating point arithmetic.
17. An image compression method according to claim 11 , wherein the singular value decomposition of the T and the T α are performed by integer arithmetic.
18. An image compression method according to claim 12 , wherein the singular value decomposition of the T and the T α are performed by integer arithmetic.
19. An image compression method according to claim 11 , wherein a known k 2 dividing process is used together with the k 2 dividing process, so that the given image matrix X is converted into a multi-divided image.
20. An image compression method according to claim 12 , wherein a known k 2 dividing process is used together with the k 2 dividing process, so that the given image matrix X is converted into a multi-divided image.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
February 2, 2007
April 17, 2012
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.